parth/smodifier
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Adding changes required for furthur editing

Browse Source
This commit is contained in:
parth aranke 2024-06-04 10:39:49 +05:30
parent 4f9a670aa7
commit c7fddcaae4
2 changed files with 14 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
Utils.py
View File

@ -59,7 +59,7 @@ class ProceesAudio():
model = BigramLanguageModel() model = BigramLanguageModel()
model.to("cpu", dtype=float) model.to("cpu", dtype=float)
model.load_state_dict(torch.load('amp_net.pth', map_location=torch.device('cpu'))) model.load_state_dict(torch.load('amp_net.pth', map_location=torch.device('cpu')))
return model.generate(torch.tensor(input_amps[:index+1]).view(1, index+1, len(input_amps[0])).to(self.device), len(input_amps)-index + 1) return model.generate(torch.tensor(input_amps[:index+1]).view(1, index+1, len(input_amps[0])).to(self.device), len(input_amps)-index - 1)
def make_smooth(self, audio, gain, prev_gain): def make_smooth(self, audio, gain, prev_gain):
smooth_index = 1000 smooth_index = 1000
@ -77,22 +77,27 @@ class ProceesAudio():
max_amp = max(max_amp, np.max(np.abs(segment))) max_amp = max(max_amp, np.max(np.abs(segment)))
x, _ = self._calculate_average_amplitude(segments=segments, sr=sr) x, _ = self._calculate_average_amplitude(segments=segments, sr=sr)
#print(x.shape)
index = int((int(index)/100)*len(x))
y = self._get_output_amps(x, index) y = self._get_output_amps(x, index)
print(len(y), len(x))
modified_segs = [] modified_segs = []
prev_gains = np.ones(10) prev_gains = np.ones(10)
max_part_amp = 0
for segment, mod in zip(segments, y[0]): for segment, mod in zip(segments, y[0]):
_, audios = split_stereo(segment=segment, max_amp=max_amp, sr=sr, num_parts=10) _, audios = split_stereo(segment=segment, max_amp=max_amp, sr=sr, num_parts=10)
final_audios = [] final_audios = []
curr_gains = [] curr_gains = []
for audio, target_amp, i in zip(audios, mod, range(10)): for audio, target_amp, i in zip(audios, mod, range(10)):
gain = (target_amp.item()/np.mean(np.abs(audio)))*max_amp gain = (target_amp.item()/np.mean(np.abs(audio)))*max_amp
if np.mean(np.abs(audio)) == 0: if np.mean(np.abs(audio)) == 0:
gain=0 gain=0
elif gain <= 50: elif gain <= 50 and gain >=1:
gain = gain/50 gain = gain/50
else: else:
gain=1 gain=1
@ -105,6 +110,9 @@ class ProceesAudio():
modified_seg = add_stereo(final_audios, len(final_audios[0]), sample_rate=sr) modified_seg = add_stereo(final_audios, len(final_audios[0]), sample_rate=sr)
modified_segs.append(modified_seg) modified_segs.append(modified_seg)
modified_segs[:index+1] = segments[:index+1]
modified_segs = np.concatenate(modified_segs) modified_segs = np.concatenate(modified_segs)
return modified_segs.astype(np.float32) return modified_segs.astype(np.float32)

4
server.py
View File

@ -64,7 +64,7 @@ def process_audio_bytes(audio_bytes, index):
pa = ProceesAudio() pa = ProceesAudio()
processed_data = pa.perform_modulation(data=data, sr=sample_rate, index=index) processed_data, part_amps = pa.perform_modulation(data=data, sr=sample_rate, index=index)
file_url = upload_to_s3(processed_data=processed_data, userId="parth") file_url = upload_to_s3(processed_data=processed_data, userId="parth")
arr_to_show = [] arr_to_show = []
@ -76,7 +76,7 @@ def process_audio_bytes(audio_bytes, index):
for i in range(len(arr_to_show)): for i in range(len(arr_to_show)):
arr_to_show[i] = float(arr_to_show[i]) arr_to_show[i] = float(arr_to_show[i])
return jsonify({"file_url": file_url, "array": arr_to_show}) return jsonify({"file_url": file_url, "array": arr_to_show, "part_amps": float(part_amps)})
@app.route('/modify', methods=['POST']) @app.route('/modify', methods=['POST'])
def modify(): def modify():